【摘要】 选取甘南藏族自治州为研究区域,以2016-2017年野外实测样方盖度数据和无人机照片提取的草地植被覆盖度数据为基础,对基于像元二分模型计算的草地植被覆盖度进行精度校正,分析并探讨无人机用于野外草地植被覆盖度调查的可行性,构建基于不同遥感数据源草地植被覆盖度的回归模型,并对模型进行精度评价.结果表明,利用无人机在草地上空一定距离(25 m)获取的照片可以多时相匹配Landsat 8等中分辨率遥感影像,动态监测野外大面积草地植被覆盖度;乘幂模型对以像元二分法计算的陆地卫星-8陆地成像仪(Landsat 8 OLI)产品的草地植被覆盖度校正效果最佳,模型的估测精度高达93.60%,在进行空间小尺度研究时模型计算精度最高;对数模型对以二分法计算的MOD13Q1产品的草地植被覆盖度校正效果最佳,模型的估测精度为91.97%;用中分辨率遥感数据Landsat 8 OLI校正低分辨率MODIS模型,修正后的模型R2=0.64,比原始的中分辨率成像光谱仪(MODIS)估测模型(R2=0.23)明显提高,在进行空间大尺度研究时该模型更适应.更多还原

【Abstract】 Gannan Tibetan Autonomous Prefecture was selected as the study area and, based on field-coverage data on the field measurements in 2016-2017 and the grassland fractional vegetation cover(FVC)data extracted from unmanned aerial vehicle(UAV) photos, the accuracy correction of grassland FVC based on the pixel binary model was used to analyze and discuss the feasibility of UAV applied in a grassland FVC survey. A regression model based on grassland FVC from different remote sensing data sources was constructed and the accuracy of the model evaluated. The results showed that the images obtained by UAV at a certain distance(25 m) over the grassland could be matched with Landsat 8 and other medium-resolution remote sensing images to dynamically monitor the grassland FVC of large-area grassland in the field. The power model had the best correction effect on the vegetation coverage of the Landsat 8 operational land imager(OLI) product, which was computed via the binary method of pixels, and the estimation precision of this model was as high as 93.60%. The model had the highest computational precision in a small-scale study of space. The logarithmic model had the best correction effect for the grassland FVC of the MOD13 Q1 product with the binary method, and the estimation accuracy of the model was 91.97%. The low resolution moderate-resolution imaging spectroradiometer(MODIS) model was corrected with medium-resolution remote sensing data Landsat 8 OLI, with the modified model R2= 0.64,which was significantly higher than the original MODIS estimation model(R2= 0.23), and was thus more suitable for a large-scale spatial research.更多还原